TiDB 3.0 Upgrade Guide
This document is targeted for users who want to upgrade from TiDB 2.0 or 2.1 to 3.0 versions, or from an earlier 3.0 version to later 3.0 versions. TiDB 3.0 is compatible with TiDB Binlog of the cluster version.
Upgrade caveat
Rolling back to 2.1.x or earlier versions after upgrading is not supported.
Before upgrading to 3.0 from 2.0.6 or earlier versions, check if there are any running DDL operations, especially time-consuming ones like
Add Index
. If there are any, wait for the DDL operations to finish before you upgrade.Parallel DDL is supported in TiDB 2.1 and later versions. Therefore, for clusters with a TiDB version earlier than 2.0.1, rolling update to TiDB 3.0 is not supported. To upgrade, you can choose either of the following two options:
- Stop the cluster and upgrade to 3.0 directly.
- Roll update to 2.0.1 or later 2.0.x versions, and then roll update to the 3.0 version.
Step 1: Install Ansible and dependencies on the Control Machine
TiDB Ansible release-3.0 depends on Ansible 2.4.2 ~ 2.7.11 (2.4.2 ≦ ansible ≦ 2.7.11
, Ansible 2.7.11 recommended) and the Python modules of jinja2 ≧ 2.9.6
and jmespath ≧ 0.9.0
.
To make it easy to manage dependencies, use pip
to install Ansible and its dependencies. For details, see Install Ansible and its dependencies on the Control Machine. For offline environment, see Install Ansible and its dependencies offline on the Control Machine.
After the installation is finished, you can view the version information using the following command:
ansible --version
ansible 2.7.11
pip show jinja2
Name: Jinja2
Version: 2.10
pip show jmespath
Name: jmespath
Version: 0.9.0
Step 2: Download TiDB Ansible to the Control Machine
Log in to the Control Machine using the
tidb
user account and enter the/home/tidb
directory.Back up the
tidb-ansible
folders of TiDB 2.0, 2.1, or an earlier 3.0 version using the following command:mv tidb-ansible tidb-ansible-bakDownload the tidb-ansible with the tag corresponding to TiDB 3.0. For more details, See Download TiDB Ansible to the Control Machine. The default folder name is
tidb-ansible
.git clone -b $tag https://github.com/pingcap/tidb-ansible.git
Step 3: Edit the inventory.ini
file and the configuration file
Log in to the Control Machine using the tidb
user account and enter the /home/tidb/tidb-ansible
directory.
Edit the inventory.ini
file
Edit the inventory.ini
file. For IP information, see the /home/tidb/tidb-ansible-bak/inventory.ini
backup file.
Make sure that
ansible_user
is the normal user. For unified privilege management, remote installation using the root user is no longer supported. The default configuration uses thetidb
user as the SSH remote user and the program running user.## Connection # ssh via normal user ansible_user = tidbYou can refer to How to configure SSH mutual trust and sudo rules on the Control Machine to automatically configure the mutual trust among hosts.
Keep the
process_supervision
variable consistent with that in the previous version. It is recommended to usesystemd
by default.# process supervision, [systemd, supervise] process_supervision = systemdIf you need to modify this variable, see How to modify the supervision method of a process from
supervise
tosystemd
. Before you upgrade, first use the/home/tidb/tidb-ansible-bak/
backup branch to modify the supervision method of a process.
Edit the configuration file of TiDB cluster components
If you have previously customized the configuration file of TiDB cluster components, refer to the backup file to modify the corresponding configuration file in the /home/tidb/tidb-ansible/conf
directory.
Note the following parameter changes:
In the TiKV configuration,
end-point-concurrency
is changed to three parameters:high-concurrency
,normal-concurrency
andlow-concurrency
.readpool: coprocessor: # Notice: if CPU_NUM > 8, default thread pool size for coprocessors # will be set to CPU_NUM * 0.8. # high-concurrency: 8 # normal-concurrency: 8 # low-concurrency: 8Recommended configuration: the number of TiKV instances * the parameter value = the number of CPU cores * 0.8.
In the TiKV configuration, the
block-cache-size
parameter of different CFs is changed toblock-cache
.storage: block-cache: capacity: "1GB"Recommended configuration:
capacity
= MEM_TOTAL * 0.5 / the number of TiKV instances.In the TiKV configuration, you need to configure the
tikv_status_port
port for the multiple instances on a single machine scenario. Before you configure it, check whether a port conflict exists.[tikv_servers] TiKV1-1 ansible_host=172.16.10.4 deploy_dir=/data1/deploy tikv_port=20171 tikv_status_port=20181 labels="host=tikv1" TiKV1-2 ansible_host=172.16.10.4 deploy_dir=/data2/deploy tikv_port=20172 tikv_status_port=20182 labels="host=tikv1" TiKV2-1 ansible_host=172.16.10.5 deploy_dir=/data1/deploy tikv_port=20171 tikv_status_port=20181 labels="host=tikv2" TiKV2-2 ansible_host=172.16.10.5 deploy_dir=/data2/deploy tikv_port=20172 tikv_status_port=20182 labels="host=tikv2" TiKV3-1 ansible_host=172.16.10.6 deploy_dir=/data1/deploy tikv_port=20171 tikv_status_port=20181 labels="host=tikv3" TiKV3-2 ansible_host=172.16.10.6 deploy_dir=/data2/deploy tikv_port=20172 tikv_status_port=20182 labels="host=tikv3"
Step 4: Download TiDB 3.0 binary to the Control Machine
Make sure that tidb_version = v3.0.x
in the tidb-ansible/inventory.ini
file, and then run the following command to download TiDB 3.0 binary to the Control Machine:
ansible-playbook local_prepare.yml
Step 5: Perform a rolling update to TiDB cluster components
If the
process_supervision
variable uses the defaultsystemd
parameter, perform a rolling update to the TiDB cluster using the following command corresponding to your current TiDB cluster version.When the TiDB cluster version < 3.0.0, use
excessive_rolling_update.yml
.ansible-playbook excessive_rolling_update.ymlWhen the TiDB cluster version ≧ 3.0.0, use
rolling_update.yml
for both rolling updates and daily rolling restarts.ansible-playbook rolling_update.yml
If the
process_supervision
variable uses thesupervise
parameter, perform a rolling update to the TiDB cluster usingrolling_update.yml
, no matter what version the current TiDB cluster is.ansible-playbook rolling_update.yml
Step 6: Perform a rolling update to TiDB monitoring components
ansible-playbook rolling_update_monitor.yml